Intelligent detection method and system for support turning

Abstract

The invention relates to the technical field of image processing, in particular to an intelligent detection method and system for support turning. The method comprises the steps that sensor signals of the support in the turning machining process are obtained in real time; constructing a sensor topological graph; extracting real-time time sequence process features reflecting the dynamic response of the tool-workpiece-machine tool system in the sensor topological graph; inputting the real-time time sequence process features into a trained multi-task prediction model, and outputting prediction deviation values of a plurality of geometric dimensions of the support and a prediction surface defect mask pattern; and if all the predicted deviation values are within the corresponding tolerance threshold values and the ratio of the defect area of the predicted surface defect mask pattern to the total area of the bracket is smaller than a set area threshold value, judging that the expected quality state of the bracket is qualified. According to the scheme, the expected quality state of the support product can be comprehensively judged in real time.

Description

Technical Field

[0001] The present invention relates to the field of image processing technology, and more particularly to an intelligent detection method and system for stent turning. Background Art

[0002] As a precision implantable medical device, the geometric dimensional accuracy and surface integrity of stents are crucial to their clinical performance and service life. During the stent manufacturing process, turning is a core process to ensure critical dimensions and surface quality.

[0003] Traditional stent quality control methods primarily rely on offline testing after machining is complete, which involves spot checks or full inspections of finished workpieces using equipment such as coordinate measuring machines, industrial CT, optical profilometers, or high-power microscopes. This post-processing testing model exhibits significant lag and is unable to detect and warn of quality anomalies in real time during the machining process. This results in irreversible scrap once it is produced, leading to not only waste of materials and labor hours, but also a significant increase in production costs and a reduction in overall manufacturing efficiency. Furthermore, offline testing only obtains final quality results, making it difficult to establish a direct and precise causal relationship with dynamic changes during the machining process. Consequently, it cannot provide effective guidance for the real-time optimization and adjustment of process parameters, limiting the level of intelligence and automation in the stent manufacturing process.

[0004] To overcome the drawbacks of offline monitoring, academia and industry have begun researching online monitoring technologies for machining processes based on sensor signals. These technologies typically deploy force, vibration, acoustic emission, power, or temperature sensors on machine tools, attempting to indirectly determine machining status by analyzing the signals collected in real time. However, existing online monitoring technologies still have many limitations, as follows: At the information fusion level, most methods perform independent analysis or simple weighted fusion of sensor signals from different physical sources, ignoring the complex physical coupling relationship and dynamic interaction between the various components within the tool-workpiece-machine tool system, resulting in the extracted features being unable to comprehensively and accurately characterize the overall dynamic response of the system.

[0005] When modeling the correlation between features and quality, existing technologies often use traditional signal processing methods combined with shallow machine learning models. This makes it difficult to extract deep features that strongly correlate with final part geometry deviations and microscopic surface defects from high-dimensional, nonlinear sensor signals. Furthermore, most prediction models focus on a single quality metric (such as surface roughness or tool wear), failing to simultaneously and comprehensively predict and evaluate both dimensional accuracy and surface defects, two core quality factors. This makes it difficult to comprehensively and real-timely determine the expected quality status of stent products. Summary of the Invention

[0006] The purpose of the present invention is to propose an intelligent detection method and system for bracket turning processing, so as to solve the problem in the prior art that it is difficult to achieve comprehensive and real-time judgment of the expected quality status of bracket products; to this end, the present invention provides solutions in the following two aspects.

[0007] In a first aspect, the present invention provides an intelligent detection method for stent turning, comprising: Acquiring sensor signals of the bracket during turning in real time, wherein the sensor signals include a triaxial vibration signal, an acoustic emission signal, and a spindle load signal; Constructing a sensor topology graph, wherein the nodes of the sensor topology graph are any of the in-situ sensor signals, and the connection relationship between the nodes is the coupling strength between the corresponding two signals; Extract the real-time time series process characteristics reflecting the dynamic response of the tool-workpiece-machine system from the sensor topology map; Inputting the real-time time series process features into a trained multi-task prediction model, and outputting predicted deviation values ​​of multiple geometric dimensions of the bracket and a predicted surface defect mask map; If all predicted deviation values ​​are within the corresponding tolerance threshold, and the ratio of the defect area of ​​the predicted surface defect mask image to the total area of ​​the bracket is less than the set area threshold, the expected quality status of the bracket is determined to be qualified.

[0008] In the above scheme, a graph structure is constructed from lathe sensor signals during the machining of a stent. Real-time time series process features reflecting the dynamic response of the tool-workpiece-machine system are extracted from these features. These real-time time series process features are then predicted using a multi-task prediction model to obtain predicted deviation values ​​for multiple stent geometric dimensions and a predicted surface defect mask. This allows for an assessment of the stent's expected quality status. This enables a comprehensive, real-time assessment of the stent's expected quality status.

[0009] Optionally, the multiple geometric dimensions include: the maximum outer diameter of the bracket body, the inner diameter of the bracket center through hole, and the flatness of the end surface of the bracket mounting flange.

[0010] The key geometric dimensions of the bracket are selected to provide data support for subsequent bracket testing.

[0011] Optionally, the multi-task prediction model includes a decoder and an encoder, the decoder is a long short-term memory network model, and the encoder includes two branches. The first branch is composed of multiple connection layers to output prediction deviation values ​​of multiple geometric dimensions, and the loss function is a mean square error loss function; the second branch adopts a U-Net network structure, and the loss function is a combination of Dice loss and cross entropy loss.

[0012] Optionally, the training process of the multi-task prediction model is: Acquire a training set; the training set includes time series process characteristics of sensor signals of multiple stent samples during processing and corresponding labels; the labels include deviation values ​​of multiple geometric dimensions and surface defect maps; The multi-task prediction model is trained using the training set, and the losses of the first branch and the second branch are calculated respectively. The parameters of the network prediction model are adjusted using the gradient descent algorithm until the loss value between the predicted value and the label output by the two branches is less than the threshold or the number of training times reaches the set number. The training is stopped and the trained network prediction model is obtained.

[0013] By training the multi-task prediction model, the robustness of the multi-task prediction model can be improved.

[0014] Optionally, the process of obtaining the deviation values ​​of the multiple geometric dimensions and the surface defect map is: Acquire three-dimensional point cloud data of the stent sample; align and register the three-dimensional point cloud data with a standard CAD digital model of the stent sample to calculate deviation values ​​of various geometric dimensions; Point clouds whose geometric dimension deviation values ​​are greater than the set deviation threshold are marked as surface defects to obtain a surface defect map of the three-dimensional point cloud data.

[0015] Label the stent samples to provide supervision signals for the training of the multi-task prediction model and drive the update of the model parameters.

[0016] Optionally, the real-time time series process features are obtained using a spatiotemporal graph convolutional network model.

[0017] Optionally, the sensor topology map is obtained by processing the sensor signal using a sensor topology map neural network.

[0018] Optionally, the connection relationship between the nodes includes spatial edges, functional edges and dynamic edges; the spatial edge is the reciprocal of the Euclidean distance between sensors in the spatial layout; the functional edge is the causal or correlation strength between the sensor signal time series corresponding to two different nodes analyzed by Granger causality test or mutual information algorithm; the dynamic edge is based on the signal similarity between the sensor signal time series corresponding to the two nodes.

[0019] The above-mentioned connection relationship can quantify the strength of the relationship between nodes, thereby driving more refined graph algorithm behavior.

[0020] Optionally, the step of acquiring the sensor signal includes: installing at least one vibration sensor on the machine tool tool holder to collect multi-axial vibration signals; installing an acoustic emission sensor on the workpiece fixture or tool holder near the cutting area to collect acoustic emission signals; and acquiring a spindle load signal from the machine tool control system, wherein the spindle load signal is a current signal.

[0021] In a second aspect, an intelligent detection system for stent turning processing includes: processor; The memory stores computer instructions for intelligent detection of bracket turning processing. When the computer instructions are executed by the processor, the system executes the above-mentioned intelligent detection method for bracket turning processing.

[0022] The beneficial effects of the present invention are: This invention utilizes a sensor topology graph neural network to explicitly model the coupling relationships between different sensors as a graph structure. This allows for the extraction of real-time time series process features from multi-source sensor signals that more comprehensively and accurately reflect the overall state of the tool-workpiece-machine system. This overcomes the drawback of traditional methods that simply fuse signals and lose critical contextual information. Furthermore, by constructing a multi-task prediction model that correlates sensor signals with the three-dimensional point cloud data of the completed bracket, this approach addresses the difficulty of synchronously and comprehensively evaluating multi-dimensional quality, a challenge faced by existing technologies. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Figure 1 A flowchart schematically illustrates the steps of an intelligent detection method for stent turning processing in this embodiment; Figure 2 The structural block diagram of an intelligent detection system for stent turning processing in this embodiment is schematically shown. DETAILED DESCRIPTION

[0024] The technical solutions in the embodiments of the present invention will be described clearly and completely below with reference to the accompanying drawings in the embodiments of the present invention.

[0025] like Figure 1 As shown, an intelligent detection method for stent turning in this embodiment includes the following steps: Step S1 , acquiring sensor signals of vibration, acoustic emission and spindle load of the bracket during turning processing in real time.

[0026] Specifically, the sensor signals include vibration signals in the X, Y, and Z directions collected by at least one three-axis vibration sensor installed on the machine tool tool holder; high-frequency stress wave signals collected by an acoustic emission sensor installed on the workpiece fixture or tool holder near the cutting area; and current signals reflecting spindle load changes (i.e., spindle load signals) collected by the current sensor integrated on the machine tool spindle drive motor.

[0027] In this embodiment, a high-speed data acquisition card, such as NI's PXI series acquisition card, is used to synchronously acquire vibration and load signals at a sampling frequency of not less than 50 kHz, and to acquire acoustic emission signals at a sampling frequency of not less than 2 MHz, forming a continuous, multi-channel time series data stream, which is then transmitted in real time to an edge computing unit or industrial computer for subsequent processing.

[0028] The sampling frequency of the triaxial accelerometer is typically set to 25.6 kHz to capture key dynamic information during machining, such as the chatter characteristic frequency. The acoustic emission sensor collects signals with a frequency range of 100 to 400 kHz, accurately capturing the instantaneous signals of tiny tool chipping or chip morphology changes. The spindle motor load data (current signal) is read 100 times per second from the servo drive of the machine tool's CNC system, reflecting the macroscopic trends in cutting force. For example, as the tool gradually wears, the spindle load will show a slowly increasing trend.

[0029] Step S2: construct a sensor topology map, where the nodes of the sensor topology map are any of the in-situ sensor signals, and the connection relationship between the nodes is the physical coupling strength between the corresponding two signals; extract the real-time time series process characteristics reflecting the dynamic response of the tool-workpiece-machine tool system in the sensor topology map.

[0030] Specifically, a sensor topology graph is constructed, in which each node represents a sensor signal, such as the X-axis vibration signal, the Y-axis vibration signal, the acoustic emission channel, etc.; the connection relationship between the nodes represents the coupling strength between different signal sources.

[0031] For example, a multi-source sensor system is abstracted as a cyber-physical graph. Each node in the graph represents an independent signal flow. For example, a system may have five nodes, representing X-axis vibration, Y-axis vibration, Z-axis vibration, acoustic emission signal, and spindle load signal. The characteristics of each node at any moment are the raw signal data or extracted feature vectors of the corresponding sensor signal channel within the corresponding time window.

[0032] In one embodiment, the sensor signals are processed using a sensor topology graph neural network to construct a graph structure.

[0033] The weights of the edges connecting nodes can be determined empirically. For example, because the three vibration sensor nodes are physically tightly coupled, the edge weights between them are set higher, such as 0.9. The edge weights between the vibration sensor nodes and the spindle motor load node are set based on the force transmission path within the machine tool structure. The longer the path and the more complex the connection, the lower the weight, such as 0.4.

[0034] As another implementation, a Granger causality test or a mutual information algorithm may be used to analyze the causal or correlation strength between different sensor signal time series, and the calculated quantized values ​​may be used as weights of edges in the graph.

[0035] The above-mentioned sensor topology neural network can more focusedly integrate sensor information with stronger physical correlation when aggregating sensor information, thereby gaining a deeper understanding of the complex evolution of the processing state.

[0036] Among them, the real-time time series process features are obtained by inputting the graph structure into the spatiotemporal graph convolutional network STGCN model.

[0037] Among them, the spatiotemporal graph convolutional network model (STGCN) propagates and aggregates information in the sensor space dimension through graph convolution operations, and captures the temporal evolution law of each sensor signal through temporal convolution operations to output low-dimensional, high-information-density time series process characteristics that integrate the physical coupling relationship of the system.

[0038] Step S3: input the real-time time series process features into a pre-trained multi-task prediction model, and output predicted deviation values ​​of multiple geometric dimensions of the bracket and a predicted surface defect mask image.

[0039] The multi-task prediction model uses an encoder-decoder architecture. The encoder uses a long short-term memory (LSTM) network model to encode the extracted time series process features into a fixed-length context vector that contains comprehensive state information for the entire processing process.

[0040] The decoder consists of two branches. The first branch is the size deviation regression head, which is composed of multiple fully connected layers. Its output is three or more continuous values, corresponding to the size deviations of the inner diameter, outer diameter and rib width of the stent, respectively. The first branch adopts the mean square error loss function.

[0041] The second branch is the surface defect segmentation head, which adopts the U-Net network structure and its loss function is a combination of Dice loss and cross entropy loss. The context vector is decoded and upsampled to output a defect mask image with the same size as the two-dimensional image unfolded on the stent surface. The value of each pixel in the image represents the probability of a defect at the said location, and the category of the defect, such as scratches, burrs or pits, is also output.

[0042] Among them, the pre-training process of the multi-task prediction model is: First, obtain the training set.

[0043] The training set includes historical sensor signals of multiple stent samples and corresponding labels.

[0044] The historical sensor signals are the operating status data of the lathe during the processing; the labels are the deviation values ​​of the geometric dimensions of the bracket and the surface defect map of the bracket.

[0045] Among them, multiple scaffold samples can include 200 scaffold samples in one sample batch.

[0046] The geometric dimensions of the aforementioned bracket include the outer diameter of the bracket body, the inner diameter of the bracket's central through-hole, and the flatness of the bracket's mounting flange. The maximum outer diameter determines whether the bracket can be successfully installed in its corresponding housing or cavity. The inner diameter of the bracket's central through-hole is used to mount a rotating shaft or fixing bolts, and the accuracy of its inner diameter directly affects the stability and dynamic balance of the connection. The flatness of the mounting flange ensures a tight fit between the bracket and other components and uniform force distribution. If the predicted flatness exceeds the required tolerance of 0.03 mm, stress concentration and micro-gaps will occur after assembly, potentially causing vibration and fatigue fracture over long-term operation.

[0047] Defect results include defect categories, which include surface scratches caused by chip entanglement, chatter marks on the machined surface of the bracket caused by tool wear, or surface burrs caused by the shedding of built-up edge. Typically, chatter marks on the machined surface caused by tool wear appear as regular ripples on the workpiece surface. This is due to increased wear on the tool's flank, which reduces the rigidity of the cutting system and produces high-frequency chatter. For example, when the vibration sensor detects that the main vibration frequency has deviated from the normal 200 Hz and stabilized at 650 Hz, it often indicates that chatter marks have formed on the surface, and its roughness value may deteriorate from Ra0.8 microns to Ra3.2 microns. Surface burrs caused by the shedding of built-up edge are caused by the formation of tiny welds in the chip material under high temperature and pressure near the tool tip. These welds are unstable and periodically fall off, with some of them embedding into the newly machined surface to form hard spots or burrs. This is usually associated with small fluctuations in the spindle load signal.

[0048] Among them, the deviation value of the geometric dimension of the bracket is obtained by using a three-dimensional measuring device (such as a high-precision laser three-dimensional scanner) to obtain the three-dimensional point cloud data of each bracket sample; the three-dimensional point cloud data is aligned and registered with the standard CAD digital model of the bracket, and the absolute value of the difference between the actual value and the design value of the geometric dimension is automatically calculated by the software as the deviation value of the corresponding geometric dimension.

[0049] For example, after the three-dimensional point cloud data is aligned with the standard CAD digital model of the bracket, the deviation value of the geometric dimension can be obtained, such as the outer diameter deviation of 0.08 mm, the flatness deviation of 0.04 mm and other specific numerical labels.

[0050] Among them, the surface defect map of the bracket is to mark the point cloud whose calculated geometric dimension deviation value is greater than the set deviation threshold as a surface defect, so as to obtain a binary defect map of the bracket sample. The pixel value of the marked surface defect is 1, and the pixel value of the non-defect is 0.

[0051] The above-mentioned deviation threshold value may be set to 20 micrometers. Of course, as other implementations, it can also be determined according to actual conditions.

[0052] It should be noted that when obtaining the training set, different cutting parameter combinations were introduced during the machining of the bracket samples to cover a range of quality conditions, from acceptable to unacceptable. For each bracket sample, the corresponding multi-channel sensor data was fully recorded throughout the entire process, from blank clamping to finished unloading, and uniquely associated with the corresponding bracket sample.

[0053] After obtaining the trained multi-task prediction model, the real-time time series process features are input into the trained multi-task prediction model to obtain a prediction deviation value and a predicted surface defect mask image.

[0054] Step S4: determining the expected quality status of the stent in real time based on the predicted deviation values ​​of multiple geometric dimensions of the stent and the predicted surface defect mask image.

[0055] Specifically, the evaluation process of the expected quality status of the stent in this embodiment is as follows: If the predicted deviation values ​​of all geometric dimensions are within the corresponding tolerance thresholds and the predicted defects do not meet the preset defect judgment criteria, the expected quality status of the bracket is judged to be qualified; conversely, if the predicted deviation value of any geometric dimension exceeds the tolerance threshold or the predicted defect meets the preset defect judgment criteria, it is judged to be unqualified, and early warning information can be given based on the specific predicted deviation value or defect result.

[0056] The above-mentioned preset defect determination criteria include: the ratio of the defect area of ​​the predicted surface defect mask to the total area of ​​the stent is greater than or equal to a preset area threshold. The total area of ​​the stent is the sum of the number of pixels in the captured image of the stent.

[0057] The preset area threshold value may vary depending on the defect. For example, the preset threshold value for surface scratches and burrs is 0.2%.

[0058] The above tolerance thresholds include an inner diameter tolerance threshold, an outer diameter tolerance threshold, and a rib width tolerance threshold, and specific values ​​may be 5 microns, 5 microns, and 3 microns, respectively.

[0059] By establishing a double-check quality judgment logic, the quality of the stent can be accurately evaluated.

[0060] The solution of the present invention can predict two core quality indicators, namely the deviation value of the geometric dimensions of the stent and the surface defects, in parallel from the processing data of the processing process, thereby being able to timely and effectively evaluate the expected quality of the stent.

[0061] The present invention also provides an intelligent detection system for bracket turning. Figure 2 As shown, the system includes a processor and a memory, wherein the memory stores computer program instructions. When the computer program instructions are executed by the processor, the intelligent detection method for bracket turning processing according to the present invention is implemented.

[0062] The system further includes other components well known to those skilled in the art, such as a communication bus and a communication interface. The configuration and functions of these components are known in the art and thus will not be described in detail here.

[0063] In the present invention, the aforementioned memory can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. For example, the computer-readable storage medium can be any suitable magnetic or magneto-optical storage medium, such as Resistive Random Access Memory (RRAM), Dynamic Random Access Memory (DRAM), Static Random Access Memory (SRAM), Enhanced Dynamic Random Access Memory (EDRAM), High-Bandwidth Memory (HBM), Hybrid Memory Cube (HMC), etc., or any other medium that can be used to store the required information and can be accessed by an application, module, or both. Any such computer storage medium can be part of, accessible to, or connectable to a device. Any application or module described in the present invention can be implemented by computer-readable / executable instructions stored or otherwise retained by such a computer-readable medium.

[0064] In the description of this specification, “a plurality of” means at least two, for example, two, three or more, etc., unless otherwise clearly defined.

[0065] Although this specification has shown and described several embodiments of the present invention, it is obvious to those skilled in the art that such embodiments are provided by way of example only. Those skilled in the art will conceive of many modifications, changes and substitutions without departing from the idea and spirit of the present invention.

Claims

1. An intelligent detection method for stent turning, characterized in that: The following steps are involved: Acquiring sensor signals of the bracket during turning in real time, wherein the sensor signals include a triaxial vibration signal, an acoustic emission signal, and a spindle load signal; Constructing a sensor topology graph, wherein the nodes of the sensor topology graph are any of the in-situ sensor signals, and the connection relationship between the nodes is the coupling strength between the corresponding two signals; Extract the real-time time series process characteristics reflecting the dynamic response of the tool-workpiece-machine system from the sensor topology map; Inputting the real-time time series process features into a trained multi-task prediction model, and outputting predicted deviation values ​​of multiple geometric dimensions of the bracket and a predicted surface defect mask map; If all predicted deviation values ​​are within the corresponding tolerance threshold, and the ratio of the defect area of ​​the predicted surface defect mask image to the total area of ​​the bracket is less than the set area threshold, the expected quality status of the bracket is determined to be qualified.

2. The intelligent detection method for stent turning according to claim 1, characterized in that: The multiple geometric dimensions include: the maximum outer diameter of the bracket body, the inner diameter of the bracket center through hole, and the flatness of the end surface of the bracket mounting flange.

3. The intelligent detection method for stent turning according to claim 1, characterized in that: The multi-task prediction model includes a decoder and an encoder. The decoder is a long short-term memory network model. The encoder includes two branches. The first branch is composed of multiple connection layers to output prediction deviation values ​​of multiple geometric dimensions, and the loss function is a mean square error loss function; the second branch adopts a U-Net network structure, and the loss function is a combination of Dice loss and cross entropy loss.

4. The intelligent detection method for stent turning according to claim 3, characterized in that: The training process of the multi-task prediction model is as follows: Acquire a training set; the training set includes time series process characteristics of sensor signals of multiple stent samples during processing and corresponding labels; the labels include deviation values ​​of multiple geometric dimensions and surface defect maps; The multi-task prediction model is trained using the training set, and the losses of the first branch and the second branch are calculated respectively. The parameters of the network prediction model are adjusted using the gradient descent algorithm until the loss value between the predicted value and the label output by the two branches is less than the threshold or the number of training times reaches the set number. The training is stopped and the trained network prediction model is obtained.

5. The intelligent detection method for stent turning according to claim 4, characterized in that: The process of obtaining the deviation values ​​of the multiple geometric dimensions and the surface defect map is as follows: Acquire three-dimensional point cloud data of the stent sample; align and register the three-dimensional point cloud data with a standard CAD digital model of the stent sample to calculate deviation values ​​of various geometric dimensions; Point clouds whose geometric dimension deviation values ​​are greater than the set deviation threshold are marked as surface defects to obtain a surface defect map of the three-dimensional point cloud data.

6. The intelligent detection method for stent turning according to claim 1, characterized in that: The real-time time series process features are obtained using a spatiotemporal graph convolutional network model.

7. The intelligent detection method for stent turning according to claim 1, characterized in that: The sensor topology map is obtained by processing the sensor signal using a sensor topology map neural network.

8. The intelligent detection method for stent turning according to claim 7, characterized in that: The connection relationship between the nodes includes spatial edges, functional edges and dynamic edges; the spatial edge is the inverse of the Euclidean distance between sensors in the spatial layout; the functional edge is the causal or correlation strength between the sensor signal time series corresponding to two different nodes using Granger causality test or mutual information algorithm; the dynamic edge is based on the signal similarity between the sensor signal time series corresponding to two nodes.

9. The intelligent detection method for stent turning according to claim 1, characterized in that: The sensor signal acquisition step includes: installing at least one vibration sensor on the machine tool tool holder to collect multi-axial vibration signals; installing an acoustic emission sensor on a workpiece fixture or tool holder near the cutting area to collect acoustic emission signals; and obtaining a spindle load signal from the machine tool control system, wherein the spindle load signal is a current signal.

10. An intelligent detection system for stent turning, characterized in that: include: processor; A memory storing computer instructions for intelligent detection of bracket turning processing, wherein when the computer instructions are executed by the processor, the system executes an intelligent detection method for bracket turning processing according to any one of claims 1 to 9.

Images